Abstract: With the continuous development of revolution of energy consumption and supply in China, the development of thermal power plants are facing a severe situation that thermal power enterprises need to improve their competitiveness in new-generation power systems. The power generation mode based on cyber physical system (CPS) was proposed based on the general structure and application mode of the CPS for the working process modeling and operation optimization of cold-end system of indirect air-cooling power units. A data-driven mechanism analysis method was introduced to build the cold-end CPS model and guide operation. The history operation data of case unit was illustrated to modify the characteristic curves of stage efficiency for the off-design calculation of turbo-units. A back propagation neural network (BPNN) was built for forecasting the air cooling tower outlet water temperature. The optimal pump frequency was obtained by coupling simulation. The case study shows that the utilization of CPS-based modeling and optimizing strategy results in a 0.9g/(kW·h) decrease in net coal consumption rate so as to improve the unit economic performance.